Determinants of capital structure in the hospitality industry: Impact of clustering and seasonality on debt and liquidity

Volatile seasonal demand and geographic clustering of firms are two important factors affecting the capital structure of hospitality firms. In this paper, we investigate the determinants of capital structure of hospitality firms with emphasis on the effects of seasonality and firm clustering. A fixed-effects panel data model was estimated using data on all hospitality firms in Norway from 2008 to 2018. Our empirical findings reveal that the seasonality created by foreign tourists increases the share of long-term debt in firms’ capital structure. Further, the clustering of hospitality firms in a region enhances firms’ reliance on short-term debt. Firms’ liquidity is negatively associated with the degree of clustering, suggesting that greater competition drains cash and short-term debt serves as a substitute. These findings have important implications for financial management by firms in the hospitality industry as the degrees of seasonality and clustering significantly affect the financing of assets and liquidity management.


Introduction
Early capital structure theories (Modigliani & Miller, 1958, 1963Jensen & Meckling, 1976;DeAngelo & Masulis, 1980;Myers, 1977Myers, , 1984Myers & Majluf, 1984) assume that firmlevel financing structure is independent of industry characteristics. However, given the considerable differences across many industries in terms of the operations of their firms and supply and demand throughout the operating cycle, it is unlikely that the capital structure theories are applicable to all industries, not least the hospitality industry, consisting primarily of hotels and restaurants 1 . As a result, the capital structure of hotel and restaurant firms may well depend on industry characteristics. Importantly, while the hospitality industry is unique in many respects, volatile seasonal demand and the geographic clustering of firms may well be two of the more important factors affecting the capital structure of its firms. First, tourism seasonality creates imbalances in the number of tourists, hotel overnight stays, and expenditures (Barros & Sousa, 2019), resulting in underutilized assets out of season if the design of firm capacity is to meet peak demand.
For example, recent work by Zhang et al. (2020) provides evidence that seasonality affects the firm-level profitability of hotels and restaurants. Furthermore, it is also likely that seasonality influences capital structure beyond its effect on profitability, given that investors and creditors may view a firm facing cyclical demand as riskier.
Second, industry clustering, defined as the geographic concentration of interrelated firms (Porter, 1998), may also have a role to play in capital structure. Firms within a cluster can benefit from their advantages in information and knowledge sharing, access to public goods, and from synergies arising from specialization within the cluster (Porter, 1990). Accordingly, clustering should positively affect firm performance. However, increased rivalry within the cluster can also negatively affect the economic performance of its constituent firms (Marco-Lajara et al., 2014). This is particularly true for the service industry where demand is largely constrained within an area, and product differentiation (e.g., among hotels and among restaurants) is relatively small compared to that in the manufacturing industry. Consequently, hospitality firms within a cluster may face high competition when it comes to both attracting tourists and capital.
Moreover, if the available menu of financing options for hospitality firms depends on industry characteristics, these characteristics become important input variables in the financial management process. They may then influence whether firms finance new investments with long-and/or short-term debt and/or equity. Given that the impact of seasonality and clustering on liquidity is an important consideration for firms' financial managers, the results of our econometric analysis are important for not only academic inquiry into capital structure, but also the day-to-day operations of firms in the hospitality industry. To the best of our knowledge, there has been hitherto no comparable investigation in the literature into how tourism seasonality and clustering influence liquidity in the hospitality industry.
For the empirical analysis, we have used the Norwegian hospitality industry data. Norway is a high-cost destination given the prevailing level of prices (Xie & Tveterås, 2020a).
However, the destination's price competitiveness has greatly improved in the last few years following the significant depreciation of the Norwegian krone (NOK). As discussed in Xie & Tveterås (2020a, b), the Norwegian economy is highly dependent on oil exports, and thus the oil price crisis in 2014 has dragged down the value of the NOK. This has made it cheaper for foreigners to visit Norway and more expensive for Norwegians to travel abroad, resulting in a boom in the Norwegian tourism industry. The annual growth rate measured by the number of hotel overnight stays was 1.9% and 4.0% before and after the oil crisis, respectively (Xie & Tveterås, 2020a). Provided by the econometric model with both firm, year, and regional fixed effects, our results indicate that while seasonal demand from international tourists increases a hospitality firm's long-term debt, a high degree of clustering in a region makes a firm more reliant on shortterm debt. One reason could be that liquidity is lower for firms located in a region with a denser clustering, indicating that competition drains cash and short-term debt serves as a substitute. This potentially increases default risk of the firms within a cluster.
Overall, the results suggest that we should augment the classical theory of capital structure by incorporating industry characteristics to make the theory applicable to the hospitality industry. An important practical implication is that hospitality firms should adapt their financial management process according to the degree of seasonal tourism concentration and the cluster effect in their region, as we demonstrate that these factors significantly influence the menu of financing options for firms and their day-to-day liquidity needs.
This paper contributes to the literature by being the first to investigate how seasonality and firm clustering influence capital structure in the hospitality industry. To investigate how these industry specific factors influence the components of capital structure (long-and shortterm debt, total debt, and liquidity), we estimate four models specifying each financing component as a dependent variable in turn. While we especially focus on seasonality and corporate clustering, we also control for other factors that previous studies have found to affect financial structure, including profitability, the growth in sales, the share of fixed assets, firm size, and non-debt tax shields (NDTS).
The remainder of the paper is organized as follows. Section 2 provides a literature review and develops the expectations for the empirical relationships. Section 3 briefly discuss capital structure in the hospitality industry. Section 4 presents the data and defines the variables.
Section 5 presents the empirical models, and Section 6 the empirical results. Finally, Section 7 presents main the findings and conclusions.

Capital structure theory
A large body of related research has followed since Modigliani and Miller (M&M) proposed their capital structure propositions (Modigliani & Miller, 1958, 1963. Among the most influential contributions are the trade-off theory (TOT) (Kraus & Litzenberger, 1973) and the pecking order framework (POF) (Myers, 1984).
M&M Proposition I suggests that under perfect capital markets, debt does not add value to the firm. M&M Proposition II further shows that if Proposition I holds, the cost of equity increases linearly with the debt to equity ratio. Further, it demonstrates that the weighted average cost of capital is constant with an increasing debt ratio. Debt capital is usually cheaper than equity, and an increase in debt, ceteris paribus, should decrease the weighted average cost of capital (WACC). However, an increase in the cost of equity as a consequence of higher debt ratio generally offsets this effect, making the WACC independent of the amount of debt in the capital structure. Further, through their tax-adjusted capital structure proposition (Modigliani & Miller, 1963), M&M predicted that the optimal capital structure is 100% debt given that interest payments is tax deductible.
TOT extends the M&M framework by taking into account the tradeoff between the deadweight costs of bankruptcy and the tax benefits of debt. According to the TOT of capital structure, management evaluates benefits and costs of different corporate debt levels. An optimal leverage is reached when marginal benefit of debt equals its marginal cost. An important component of the expected costs is the direct and indirect costs of financial distress (Andrade & Kaplan, 1998). Kraus and Litzenberger (1973) provide an early analysis of the trade-off between deadweight loss at bankruptcy and the tax benefits of debt. Graham (2003) provides an excellent review on the effects of tax on capital structure. TOT can be classified as either static or dynamic. According to the static version of TOT, a firm chooses a leverage ratio based on a single period tradeoff between tax benefits and expected bankruptcy costs (Bradley, Jarrell, & Kim, 1984). The dynamic TOT proposes that firms move towards a target leverage but is allowed to gradually adjust over time (Kane et al., 1984;Brennan & Schwartz, 1984).
The pecking order framework (POF) of capital structure (Myers, 1984) predicts that firms finance assets in a hierarchical fashion because of adverse selection. According to this theory, both the costs of financing and the degree of information asymmetry are important capital structure determinants. In essence, POF suggests that when it comes to financing assets, internal capital is preferred, followed by debt capital, while the strategy of issuing new equity is only adopted if the other financing alternatives are not available. The important motivations for POF are adverse selection (Myers & Majluf, 1984) and agency theory (Jensen & Meckling, 1976).
The discussions above show that the theories provide conflicting results regarding optimal debt usage (Barclay & Smith, 2020). Specifically, M&M proposes 100% debt, TOT suggests a debt level where marginal debt benefit equals marginal debt cost, and POF suggests that internal capital should be preferred before debt capital. However, we believe TOT and POF better explain the observed capital structure decisions of firms than the M&M propositions based on the following reasons. First, in all industries, debt ratios are practically on average well below 100% (Hall, Hutchinson & Michaelas, 2010). Few companies keep the leverage to assets ratio above 50% for longer periods of time (DeAngelo & Roll, 2015). Second, TOT explains observed debt ratios below 100% by taking into account the expected bankruptcy costs. The assumption of perfect capital markets in the M&M propositions implies no bankruptcy cost. However, bankruptcy costs create a dependence between the cash flow distribution and capital structure. Accordingly, the assumption of no dependence between the cash flow distribution and capital structure as indicated by the zero-bankruptcy cost in the M&M propositions is therefore unrealistic.
Another important contribution to capital structure theory is the market timing theory introduced by Baker and Wurgler (2002), where corporations' capital structure "…evolves as the cumulative outcome of past attempts to time the equity market (p. 27)." According to the theory, a firm issues equity when the price to book ratio is high and repurchases equity when its stock is at low market value. Accordingly, firms with high (low) leverage are those that have raised capital when the prices of their stocks were low (high) relative to their book values.
Studying IPOs, Alti (2006) finds that market timing does not have a significant long-term effect on capital structure, only in a two-year window after listing of the firm. In our study, since none of the firms in our sample are publicly listed, we resort to testing the POF and TOT of capital structure.

Components of capital structure
In this study, we investigate several components of capital structure and their determinants. The components chosen are short-term debt (STD), long-term debt (LTD), total debt (TD), and liquidity (LIQ). STD, or current liabilities (e.g., short maturity bank loans, accounts payable, wages, and income taxes payable, and current portion of long-term debt) is the firm's financial obligations expected to mature within a year. LTD is debt that is expected to be paid off in more than one year (e.g., long-term bank loans and long-term bonds). Total debt is the sum of STD and LTD. Finally, our liquidity component is net current assets, meaning current assets (the most liquid assets in the balance sheet: e.g., cash and cash equivalents, inventory, accounts receivable, marketable securities, and other liquid assets) minus current liabilities (explained above).
Since some firms must substitute short-term debt for long-term debt if they lack success in raising longer-term debt (Chittenden et al., 1996). Small firms also need to rely on shortterm debt financing as a substitute for long-term financing (MacMillan, 1931;Bolton, 1971;Wilson, 1980;Chittenden et al., 1996), we include both STD and LTD to capture any tradeoff between the debt of different maturities. We include TD as a dependent variable to identify as such a trade-off, and we can measure the net impact of changes in STD and LTD on TD.
We might, for example, observe the following: if there is a perfect substitution between STD and LTD, TD will remain unchanged. If not, TD might be affected by an increase or decrease in either of the two total debt components. In addition, we include LIQ as a component of capital structure because we want to examine how liquid assets are affected by seasonality and clustering.
The primary purpose of STD is to ensure that the firm has cash available for day-to-day operations. Therefore, especially when the firm's revenues are insufficient to cover the operational needs, STD can be a beneficial source of capital. The main advantage of LTD over STD is that a firm has a longer time to repay the loan and can therefore finance larger investments in long-term projects. LTD is used to fund investments necessary to maintain existing capacity and also to finance expansions and new projects. Liquidity analysis is also important for firms as it is used to determine a firm's ability to pay off current debt without raising additional capital. The dependent variables are defined in section 4.2.

Determinants of capital structure
Having described the components of capital structure that will be our dependent variables, we now turn to discuss the determinants of the components of firm capital structureour independent variables. This section discusses how the conventional determinants, including profitability, growth, asset structure, size, and non-debt tax shield, affect firm capital structure following the TOT and POF theories. After that, we summarize the TOT and POF predictions of how these variables affect the determinants on capital structure. The hospitability industry specific determinants of capital structure are discussed in sections 3.1 and 3.2. The independent variables are defined in Table 3.

Profitability
Since profitable firms are more likely to take advantage of the tax shield provided by debt (Toy et al., 1974;Chittenden et al., 1996;Tang & Yang, 2007;Pacheco & Tavares, 2017; Li & Singal, 2019) and have lower expected bankruptcy costs as well, TOT therefore predicts that profitable firms use more debt in their capital structure. However, according to POF, a firm finance its capital with accumulated equity first, then debt, and as a last resort, newly issued equity. This framework predicts a negative relationship between profitability and leverage, such that a more profitable firm may need less external capital because it already generates capital internally. The two capital structure theories therefore have contradictory predictions about the relationship between profitability and debt.
Several empirical studies (e.g., Botta, 2019; Chittenden et al., 1996;Kim, 1997;Karadeniz et al. 2009) identifies a negative relationship between profitability and short-term debt. We also hypothesize a negative relationship between profitability and short-term debt, and the same relationship for profitability and long-term debt.

Growth
Proxies for firm growth are also previously used in the existing literature on capital structure Growth firms have greater investment opportunities, and as investment opportunities can increase agency problems between managers and creditors, TOT thus predicts a negative relationship between growth and debt (Myers 1977). According to POF, debt is expected to be positively associated with growth because growing firms may not have internal capital available to finance growth as they might already have been exhausted in financing previous growth.
TOT and POF therefore provide contradicting predictions. There is a mix of empirical findings when it comes to the direction of this relationship. Most studies find no relationship between growth and debt (e.g., Pacheco & Tavares, 2017;Chittenden et al., 1996), while Kim (1997) finds a negative relationship.

Asset structure
Asset structure is another important capital structure determinant (Friend & Lang, 1988 reduce the risk of bankruptcy and therefore increase the expected bankruptcy costs. TOT therefore predicts a positive relationship between asset structure and debt. According to POF, debt is preferred before new equity issues. This means that if internal capital is unavailable, new debt is preferred. Therefore, according to both TOT and POF, asset structure is expected to be positively related to the level of debt. If firms prefer to use long-term debt to finance long-term projects, we expect a positive relationship between asset structure and long-term debt. In addition, we expect a substitution effect between long-term debt and short-term debt. This means that the more collateral a firm can provide, the higher is the ratio of long-term debt to total capital. Consistent with the theoretical predictions, previous research has identified a positive relationship between asset structure and long-term debt (e.g., Botta, 2019; Pacheco & Tavares, 2017; Tang & Jang, 2007). As we define asset structure using the share of fixed assets in total assets then, ceteris paribus, the higher the share of fixed assets is, the lower is the share of short-term liquidity. Therefore, we expect a negative relationship between asset structure and liquidity.

Firm size
There is a problem of asymmetric information between borrowers (e.g., firms) and lenders (e.g., banks) in financial markets. Thus, firm size has been found to affect the availability of financing options (Titman & Wessels, 1988;Sheel, 1994;Tang & Yang, 2007;Mun & Jang, 2017). Creditors are more likely to know large firms and at the same time large firms are better able to provide detailed information to potential creditors. This leads to a finance gap between small and large firms such that small firms to an increasing extent need to rely on short-term debt financing as a substitute for long-term debt financing (MacMillan, 1931). This is consistent with the TOT prediction of positive relationship between size and debt as larger firms have lower bankruptcy probability (e.g., Ohlson, 1980), and therefore lower expected bankruptcy costs. If firm size affects long-term debt availability, we expect that large firms use long-term debt as a substitute for short-term debt. Through growth, large firms may also accumulate more liquid assets, and we therefore expect to see a positive relationship between firm size and liquidity. To give an overall picture of the debt predictions of TOT and POF, we have summarized the predictions in Table 1 below. Notes: "+" indicates a positive relationship, "-" a negative relationship, and "0" no relationship.

Capital Structure in the Hospitality Industry
Several studies have investigated the determinants of capital structure in hospitality firms.
Nevertheless, to best of our knowledge, no study has investigated the effects of the industry characteristics of seasonal concentration and firm clustering on hospitality firm's capital structure.
Kwansa and Cho (1995) find that there are significant indirect and direct bankruptcy costs in the restaurant industry, and therefore question the applicability of the M&M propositions in the industry. Sheel (1994) investigates the determinants of capital structure of hotel firms and compares hotel firms' capital structure with that of manufacturing firms. His study suggests differences in capital structure between the two industries. Likewise, Tang and Jang (2007) compare the capital structure covariates of the US lodging firms and software firms and again identify differences in capital structure between these two industries.
Although Sheel (1994) and Tang and Jang (2007) have not directly tested the importance of industry characteristics in firms' capital structure determinants, the two studies recommend the importance of industry-specific analysis of capital structure determinants.
For a detailed discussion of the determinants of capital structure in the hospitality industry, we further review the literature for the restaurant, food, and beverage sector and the lodging sector separately. In the restaurant, food, and beverage sector, Kim (1997) investigates the determinants of long-and short-term debt and the total debt ratio in the US restaurant industry using the financial data of 119 restaurants. The results indicate that firm size negatively influences long-term debt, sales profitability negatively links to total debt, and sales growth negatively influences all three measures of debt. Additionally, Dalbor and Upneja (2002) identify factors determining the use of long-term debt among the publicly listed restaurant firms in the US and find that firm size and financial distress positively influence the share of long-term debt in the capital structure and growth opportunities are negatively associated with long-term debt. The asset-light and fee-oriented strategy (ALFO), which allows for firm growth with a minimum investment in assets, is found to increase the longterm debt of restaurant firms (Li & Singal, 2019) 2 . Mun and Jang (2017) find that the US restaurant firms use debt financing to refinance debt maturing in two and three years.
Analyzing long term debt ratios from a behavioral perspective, Seo, Kim and Sharma (2017) find that overconfident CEOs in the US finance the restaurant firm using more long-term debt when facing greater growth opportunities and low cash.
In the lodging sector, Karadeniz et al. (2009)  (2019) include a separate econometric estimation for hotels and restaurants using data in an identical period. They find the effect of NDTS on LTD is positive in restaurant firms but negatively in hotel firms. In addition, the ratio of fixed assets to total assets, is negatively related to LTD for restaurant firms while not significantly affecting hotel firms. Lastly, Li and Singal (2019) find that restaurant firms with a high market-to-book ratio have more long-term debt, but the LTD of hotel firms is unaffected by the market-to-book ratio. In this paper, we will also investigate whether the determinants of capital structure are different between hotels and restaurants through the robustness tests of our econometric models.
The hospitality industry is unique in many aspects with perhaps the most important being the highly seasonal demand and clustering of firms in geographical locations. However, the most influential theories of capital structure (M&M, TOT, and POF) ( It is difficult to form expectations about the relationship between the different capital structure components due to the lack of theoretical support and empirical research. The empirical results may depend on the risk tolerance of equity holders and creditors. However, we expect that seasonality negatively affects long term debt since first, we assume creditors have low tolerance to the demand risk from volatile seasonal demand; second, seasonality may increase probability of bankruptcy due to its negative effect on firms' economic performance and investment (Lundtorp, 2001;Zhang et al., 2020) and TOT predicts that less profitable firms use less debt in their capital structure. If LTD is not readily available for firms exposed to highly seasonal demand, STD may be used as a substitute and be positively related to seasonality. The relationship between seasonality and TD depends on the degree of substitution. If STD fully substitutes LTD, then we expect no relationship between TD and seasonality. Since it is more likely that STD cannot fully substitute LTD, we expect that seasonality is negatively associated with TD. Lastly, liquidity may dry up in the low season, and therefore we expect a negative effect of seasonality on liquidity.

Clustering
Several studies on the hotel industry have documented that hotel firms benefit in terms of

Dependent Variables
As we discussed above, the components of capital structure include long-term debt (LTD), net liquid assets as a share of total assets (Chittenden et al. 1996). The motivation behind is to make the liquidity variable a capital structure measure, consistent with the measurement of the other dependent variables. Another reason why we divide capital structure variables and independent variables by total assets is that we wanted to deflate the variables by a scale proxy.
In this way we control for the effect of scale.
As discussed above, profitability, growth, asset structure, firm size, and NDTS are usually used in the literature as explanatory variables of capital structure. Following the literature, profitability is represented by return on sales (ROS), which is profit before interest payments and taxes, divided by total assets. Firm growth (GROWTH) is annual growth in sales. Asset structure is fixed assets divided by total assets. Firm size is proxied by TOTAL ASSETS from the balance sheet. NDTS is the non-debt tax shield and is depreciation divided by total assets. CLUSTERING is the number of firms in a region (See Appendix 1).

Measuring tourism seasonality.
We measure seasonal demand using the Gini index as it is the most commonly employed The Gini index for a particular region (r) in a given year (t) is computed as follows: where n is the number of months with no-zero overnight stays. n is 12 in our study since throughout our sample, there are a considerable amount of overnight stays in all months in every region every year. ,1 , ,2 … ,12 are the monthly shares of overnight stays in month k in year t in each region. ∑ w k S t,k n k=1 is the sum of the weighted monthly shares. The weights = 1, 2, 3…12, with the largest weight (12) goes to the smallest share, the second largest (11) to the month with the second-smallest share, and so on.
To investigate whether the seasonal demand of foreign and domestic tourists influences firm capital structure differently, we construct two separate Gini indices for hotel overnight stays of domestic tourists and foreign tourists, along with an aggregate index for total hotel overnight stays. Specifically, for the Gini index of domestic tourist hotel overnight stays,  destinations, and also in winter between January and February in Northern Norway for the Northern lights.   When it comes to the pairwise correlation between the dependent variables and the industry-specific independent variables, we can see from Table 4 that clustering exhibits its highest correlation with short-term debt. The correlation is of positive sign, indicating that short-term debt is higher in the denser clusters. For the Gini indices, Gini overall has its highest correlation with long-term debt. However, Gini foreign has a higher correlation with long-term debt financing than Gini domestic, indicating that it may be important to disaggregate the measure of seasonal demand when developing financial management policy.   Table 3 for variable definitions. *** , ** , and * denote significance at the 0.01, 0.05, and 0.10 level, respectively.

Empirical models 3
We sequentially present the econometric models for the components of capital structure 4 comprising LTD, STD, TD, and LIQ. Among the determinants of capital structure, seasonality 5 is represented by the Gini index as discussed above and clustering refers to the number of 6 hospitality firms in a region where a firm is located. 4  Before estimating the models, we winsorized 1% in each tail of the distribution of our 51 dependent and independent variables. 52 53 6. Empirical results 54

Models with overall Gini index 55
We start by discussing the estimation results of the four models with the overall Gini index 56 only (Equations 2-5), as presented in Table 5. 57 58 Notes: *** , ** , and * denote significance at the 0.01, 0.05, and 0.10 level, respectively. Clustered robust 60 standard errors (in brackets) are clustered at firm level. See Table 3 for variable definitions. 61 62

LTD model 63
The LTD estimation in Table 5 shows that the overall Gini index positively influences the 64 share of long-term debt in capital structure. This indicates that in the presence of seasonality, 65 creditors are more willing to finance assets than equity investors in the hospitality industry. 66 As the theories (Modigliani & Miller, 1958, 1963Jensen & Meckling, 1976;Myers, 1984) 67 assume that firm-level financing structure is independent of industry characteristics, 68 this study shows how the existing theories of capital structure does not well capture an 69 important industry-specific factor that significantly influences financial structure. Seasonality 70 is one of the most important factors in the hospitality industry in general. Thus, not considering 71 this factor can lead to erroneous financial policy decisions. 72 The estimated coefficient of CLUSTERING suggests that whether a firm is within a dense 73 cluster or not does not influence the share of long-term debt in its financial structure. As 74 shown, there is a positive association between asset structure and the share of long-term debt, 75 consistent with the findings in the literature (Chittenden et al., 1996;Tang & Jang, 2007). 76 Given that this is a proxy for collateral, this also suggests that a bank would be more willing 77 to provide a long-term loan if the collateral of a firm is high relative to its total assets. In 78 addition, the estimation results suggest that large firms, as measured by total assets, have a 79 larger share of long-term debt compared to small firms given that they may have more 80 established reputations and perhaps better communication with potential lenders. This is 81 consistent with the findings in Chittenden et al. (1996) and Dalbor and Upneja (2002), and 82 with the TOT as larger firms have a lower bankruptcy probability and therefore lower 83 expected bankruptcy costs. There is a negative but not significant relationship between NDTS 84 and long-term debt. 85 Further, we can see that growth is unrelated to the share of long-term debt, which is also 86 consistent with Chittenden et al. (1996). The explanation is that growth is not an asset 87 available for pledging as collateral in a loan application, coupled with the fact that growth is 88 risky. Lastly, the return on sales negatively influences long-term debt. This result is consistent 89 with POF but inconsistent with TOT. POF states that firms should first finance new projects 90 with internal capital. We expect that profitable companies have better opportunities to 91 accumulate more internal capital to finance their assets. The TOT predicts a negative 92 relationship between profitability and debt since profitable firms are more likely to being able 93 to take advantage of the debt tax shield and these firms have lower expected bankruptcy costs. 94 95

STD model 96
The overall Gini index cannot explain changes in short-term debt, while clustering can. The 97 coefficient estimate for clustering is significant and positive, indicating that firms within a 98 region with a large number of competitors rely more on short-term debt financing. In the 99 hospitality industry, many of the investments are long term in nature (e.g., property, plant & 100 equipment investments). We expect that firms would prefer to finance these long-term 101 investments using long-term capital. Our findings that clustering increases the need for short 102 term financing indicates that, in a dense cluster, there is not only competition for tourists, but 103 also for capital. Therefore, this result indicates that the net effect of clustering is negative 104 when it comes to financing. Marco-Lajara et al. (2014) document the negative effect of 105 clusters in their research on Spanish clusters, however, their focus is not on capital structure. 106 Given that clustering has a positive net effect on the creditor's decision to lend money, the 107 TOT makes a prediction that better matches our finding. However, it is uncertain whether the 108 cluster positively influences the creditor's decision. We leave this to future research. The 109 coefficient estimates for total assets and asset structure are both negative in the model for 110 short-term debt, again consistent with the results in Chittenden et al. (1996). This suggests 111 that for large firms and firms with larger shares of collateral, there is a substitution effect 112 between long-term and short-term debt in their capital structure. It further suggests it is easier 113 for these firms to use larger long-term loans to replace smaller but more frequent short-term 114 debts. The POF and TOT do not explicitly account for possible substitution effects between 115 LTD and STD. The significant and positive coefficient on NDTS is also most likely a result 116 of the substitution effect between LTD and STD. 117 Given that the model for long-term debt reveals that growth firms do not hold more long-118 term debt than other firms, and that the model for short-term debt indicates a significant and 119 positive relationship between growth and short-term debt, this suggests that growth firms must 120 rely more on debt that matures within a year to finance their assets. The estimated coefficient 121 for the return on sales (ROS) suggests that profitable firms use less short-term debt financing, 122 which is in accordance with the findings of Chittenden et al. (1996). The possibility then exists 123 that profitable firms use more equity financing. Our results are in line with the POF which 124 predict a positive relationship between growth and debt, however not in line with the TOT 125 which predicts a negative relationship (Table 1). 126 As shown in the model for short-term debt, the degree of clustering mainly affects short-term 146 debt. This might be one of the reasons to explain the negative effect of clustering on firm 147 liquidity. 148 The estimated results suggest that firms with more total assets also have greater liquidity. 149 The explanation here may be that firms that have grown have had more time to accumulate 150 liquid assets. Further, firms with a higher share of fixed assets have a lower share of liquid 151 assets in their capital structure: a result also found by Chittenden et al. (1996) in their cross-152 sectional study. These firms having a higher share of property, plants, and equipment may 153 have less room for liquid assets as a share of total assets. The coefficient on NDTS is 154 significant and negative, most likely as a result of the positive relationship between NDTS 155 and STD. Lastly, the result suggests that profitable firms also have more liquidity, indicating 156 that these firms invest at least part of their profits in short-term financial assets. 157 158

Models with disaggregated Gini indices 159
To investigate further the relationship between the seasonality of different tourist groups and 160 capital structure, we estimated the models (Equations 6-9) with disaggregated Gini indices: 161 one for domestic tourists and the other for foreign tourists. Table 6

172
The estimated coefficients of Gini domestic and Gini foreign provide more insights into 173 the effects of seasonality on firm capital structure. The estimation results suggest that while 174 domestic tourist demand variations do not influence the amount of long-term and total debt in 175 the financial structure, foreign tourist demand variations do. Therefore, it indicates that the 176 seasonal demand of foreign tourists rather than domestic tourists increases the reliance of 177 hospitality firms on long-term debt. 178 This result has important implications for hospitality firm financial policy, indicating that 179 if a firm focuses on foreign tourists, long-term debt may be a more attainable financing 180 alternative than equity. Although we find no relationship between overall seasonal demand 181 and STD and LIQ in Table 5, the estimated results of the decomposed Gini indices suggest a  182 positive and negative relationship between the seasonal demand of domestic tourists with firm 183 short-term debt and liquidity assets, respectively. Financing strategies may relate to the 184 differences in the seasonality of the domestic and foreign tourists. 185 As discussed by Zhang et al. (2020) and suggested by the means and standard errors for 186 Gini domestic and Gini foreign in Table 4 We estimate four separate econometric models: one for each of the main capital structure 232 components, namely long-term debt, short-term debt, total debt, and liquidity. We employ the 233 Gini index to measure seasonal demand concentration. To capture the possible different 234 impacts of tourism seasonality arising from different market segments on capital structure 235 decisions, we disaggregate the measure of seasonality in terms of tourism segments, such as 236 international and domestic tourists. Therefore, we construct three seasonality indices in total, 237 one for total tourism demand, one for domestic tourist demand, and the last for foreign tourist 238 demand. We first estimate the empirical models using only the overall Gini index, and then 239 re-estimate them by replacing this with disaggregated indices for domestic and foreign 240 tourists. 241 The estimated results suggest that firms facing higher seasonal demand rely more on long-242 term debt. The influence of seasonality on long-term debt also increases total debt for 243 hospitality firms. One explanation is that it is easier for a hospitality company to persuade 244 long-term creditors to finance assets than to raise new equity payments from investors, as 245 seasonality may add to the riskiness of firm cash flow. Creditors are also in a better position 246 to recuperate some of their investment in a firm in case of bankruptcy compared to the equity 247 holders, which only have a residual claim on the firm's assets. 248 Using disaggregated seasonality measures for foreign and domestic tourists, we found that 249 the seasonality created by foreign tourists rather than domestic tourists makes firms more 250 reliant on long-term debt capital. As discussed, the reasons are that hospitality firms need to 251 have sufficient capacity to meet the highly concentrated demand of international tourists, 252 which is more predictable. To increase capacity, firms then need to invest in facilities by 253 establishing long-term loans with lenders (e.g., banks). Long-term debt and total debt are 254 therefore expected to be high. Alternatively, the unpredictable seasonal demand of domestic 255 tourists creates a need for the firms to be able to adjust their short-term debt and liquidity at 256 short notice to cope with uncertain sales revenue. 257 Firm clustering is another important variable in the hospitality industry given that hotels 258 and restaurants often locate together in a specific geographic region. There is a long debate 259 concerning the supposedly positive spillover effects and rivalry between firms within the same 260 region (Peiró-Signes et al., 2015; Marco-Lajara et al., 2014). Our econometric results reveal 261 that the firms located within a dense cluster rely more on short-term debt and have less 262 liquidity than otherwise. Greater competition between firms within denser clusters suggests 263 that the competition in clusters drains firm liquidity, through either the liquidation of financial 264 assets or the increase in short-term debt. 265 The estimated results for the control variables are mostly consistent with existing 266 literature. Specifically, profitability decreases both long-and short-term debt and increases 267 liquidity. The expectation is that profitable firms invest some of their profits in both short-268 and long-term financial assets. Growth firms usually have higher total debt and lower liquidity 269 and firms with more collateral more easily establish long-term loans with banks. Larger firms 270 generally have larger shares of liquidity as these companies have accumulated more liquid 271 financial assets in their development. 272 Our results have significant theoretical and practical implications. Theoretically, our study 273 suggests the importance of industry characteristics in capital structure determination. The 274 hospitality industry is unique in terms of seasonal demand and co-location in small regions 275 where customer demand is naturally restricted. These features affect demand and supply in 276 the industry and are thus crucial in capital structure decisions. We therefore recommend the 277 expansion of existing theoretical models to include industrial attributes. In their current forms, 278 POF and TOT might make contradictory predictions about the relationship between 279 seasonality and debt in some industries due to the different industrial properties. In addition, 280 the POF predicts a negative relationship between clustering and debt, while we found a 281 positive relationship between seasonality and STD (and hence debt financing) in this industry. 282 As before mentioned, it is not clear how creditors evaluate the impact of tourism seasonality 283 and clustering on the profitability and insolvency risk of the hospitability firms, an issue left 284 for future research. 285 The study also has important implications for the financial management of firms in the 286 hospitality industry. For firms that rely on demand from foreign tourists, long-term debt is 287 often required to build up the facilities needed to meet demand in peak seasons. Firms should 288 also rely on short-term debt ready to cope with the more unpredictable demands of domestic 289 tourists. Another important policy implication is that firms within a denser cluster have a 290 different menu of finance options available, as they are typically more reliant on short-term 291 debt financing. As firms often finance long-term projects with long-term capital, being reliant 292 on short-term debt to finance operations in regions with denser clustering may indicate that 293 this type of financing serves a substitute for long-term debt financing. In these regions, there 294 may also be both greater competition for customers and in gaining sources of capital. Lastly, 295 our results indicate that the increased competition in these clusters drains liquidity, through 296 either the liquidation of financial assets or the increase in short-term debt. 297 One limitation of this study is that we did not consider long-term debt covenants. 298 Covenants can lower the risk of the creditor and the cost of debt for the firm and therefore 299 affect the supply and demand for long term debt. Additionally, firms' decisions of location 300 may be affected by some unobservable factors, which also affect capital structure. How to 301 control for the selection bias regarding the hotels' location is another direction for future study. 302 Finally, it would be interesting to conduct a global analysis of hospitality firms from different 303 countries and markets to see whether there are any significant differences between countries 304 and markets. That is left for future research. 305